Comparative Study of Different Statistical Distributions in Generating Noise Components for Privacy Preserving Data Clustering

Abstract

Privacy Preserving Data mining is one of the recent areas of research. Privacy preserving clustering is one of the functionalities of privacy preserving data mining. The present work makes a comparative study of different statistical distributions such as Poisson distribution, Normal distribution, Discrete Uniform distribution, Continuous Uniform distribution, and Exponential distribution in generating noise components for additive data perturbation in the context of privacy preserving data clustering. The results obtained show that the mean squared error obtained from the perturbed data generated by adding noise generated from both discrete uniform distribution and continuous uniform distribution is same as the mean squared error obtained from the original data but the order of values differ due to random selection of cluster centers.